Alpha-NOMINATE Applied to the 113th Senate

While we await for the Republicans to decide who will be their candidate for Speaker, it is instructive to look back at the 113th Senate with our new alpha-NOMINATE procedure. Recall that Alpha-NOMINATE is a new form of NOMINATE that is fully Bayesian and is meant to replace W-NOMINATE which is now about 32 years old (the multidimensional version, written by Nolan McCarty and Keith Poole is 24 years old). NOMINATE was designed by Keith Poole and Howard Rosenthal during 1982-1983. It used a random utility model with a Gaussian deterministic utility function (see pages 14 – 15 of the linked 1983 paper) and logistic error (random draws from the log of the inverse exponential). The Gaussian deterministic utility function is able to capture non-voting due to indifference and alienation.

Alpha-NOMINATE is a mixture model in which legislators’ utility functions are allowed to be a mixture of the two most commonly assumed utility functions: the quadratic function and the Gaussian function assumed by NOMINATE. The “Alpha” is a parameter estimated by Alpha-NOMINATE that varies from 0 (Quadratic Utility) to 1 (Gaussian Utility). Hence, in one dimension with Alpha = 0, Alpha-NOMINATE is identical to the popular IRT model. Thus Alpha-NOMINATE can actually test whether or not legislators’ utility functions are Quadratic or Gaussian.

Below we apply Alpha-NOMINATE to the 113th Senate. There were 657 total votes in the 113th of which 552 are scalable (at least 2.5% in the minority; that is, votes that are 97-3 to 50-50). We used the R version of Alpha-NOMINATE to perform the analysis. We used 3000 samples from a slice sampler in one dimension with a burn-in of 1000. The first graph shows the Trace and Density plots for alpha.


Click image to enlarge


The mean of Alpha is 0.9981 with a standard deviation of 0.0018 strongly indicating that the Senators’ utility functions were Gaussian.

The next plot shows the estimated ideal points for the 105 Senators who served during the 113th along with their 95% Credible Intervals. On the left, Presidential candidate Senator Bernie Sanders (I-VT) is located at -2.74 off the left edge of the plot. Near the right end are Presidential candidates Ted Cruz (R-TX) at 1.608 Rand Paul (R-KY) at 1.453. Lindsey Graham is near the left edge of the Republican Party at 1.011.

Click image to enlarge


There were very few moderates in the 113th Senate. Only Lisa Murkowski (R-AK), Susan Collins (R-ME), and Joe Manchin (D-WV) are between the two parties. Moderation is in short supply in both the Senate and the House. No wonder Republicans are finding it difficult to find someone to be Speaker.

Alpha-NOMINATE Applied to the 113th House

Following up on our previous posts, below we apply Alpha-NOMINATE to the 113th House. There were 1,202 total votes in the of which 1,021 are scalable (at least 2.5% in the minority). We used the R version of Alpha-NOMINATE to perform the analysis. We used 2000 samples from a slice sampler in one dimension with a burn-in of 1000. The first graph shows the Trace and Density plots for alpha.


Click image to enlarge


The mean of alpha was 0.99033 with a standard deviation of 0.001125 strongly indicating that the Representatives’ utility functions were Gaussian.

The next five plots show the estimated ideal points for the 443 scalable Representatives (Emerson (R-MO) only voted 11 times) and President Obama (using CQ Presidential Support votes his location is at -1.41) along with their 95% Credible Intervals. On the left, Representative Schakowsky (D-IL) is located at -2.262. Her 95% credible interval runs from -2.460 to -2.077. The five Republicans on the right end are Massie (R-KY) at 3.93 (3.77 – 4.09), Broun (R-GA) at 3.94 (3.80 – 4.10), Amash (R-MI) at 4.00 (3.85 – 4.14), Duncan (R-TN) at 4.08 (3.89 – 4.26), and all by himself, Jones (R-NC) at 9.16 (8.59 – 9.80).

Walter Jones is also the most extreme member of the 114th as of the August recess. Indeed, many of the more extreme members of the Republican caucus continued into the 114th. With the volatile issue of Planned Parenthood funding on the table there will be attempts to attach defunding language to many must-pass bills. Neither President Obama nor the Republican leaders want another shutdown but with the number of intense conservatives in the House Caucus it will be a tricky row to hoe.

Click image to enlarge



Click image to enlarge



Click image to enlarge



Click image to enlarge



Click image to enlarge


Leaving out Jones (R-NC):

The Median of the Democratic Party is -1.317 with a standard deviation of 0.0166 and the Median of the Republican Party is 0.993 with a standard deviation of 0.0166. The probabilities for the median Representative are 0.083 for Gerlach (R-PA) located at 0.455, 0.079 for Wolf (R-VA) at 0.449, 0.073 for Frelinghuysen (R-NY) at 0.478, 0.073 for Meehan (R-PA) at 0.451, and 0.071 for Dent (R-PA) at 0.433. Given the large number of Republicans concentrated around 0.45 these probabilities are no surprise.

Congressional Policy Shifts, 1879-2014

Below we use first dimension DW-NOMINATE scores, which represent legislators’ positions along the familiar ideological (liberal-conservative) spectrum, with lower (negative) scores indicating greater liberalism, and higher (positive) scores denoting conservatism, to plot the chamber means for legislators and winning outcomes on roll calls for Congresses 46 to 113 (2014). These are updates to Figure 4.1 (page 60) in Poole and Rosenthal’s Congress: A Political-Economic History of Roll Call Voting and Figure 4.1 (page 80) of Ideology and Congress.

We see that the overall chamber means (plotted with red squares) remain mostly stable over time, a reflection of a competitive two-party system. Of the two chambers, the House mean has shifted more to the right in the period following the 104th Congress (the 1994 “Republican Revolution”). However, the position of the mean winning coordinate in each chamber has proved much more volatile, particularly in recent Congresses. This reflects the frequency of party-line votes between rival partisan coalitions that have moved steadily apart in recent decades.

Consequently, the mean winning coordinate–-which is an approximation of the ideological location of policies enacted in the chamber–-has diverged from the overall mean chamber score in both chambers: to the left under Democratic control, to the right under Republican control. Indeed, the mean winning coordinate in the 111th Senate (a session in which the number of Senate Democrats fluctuated between 57 and a supermajority of 60) was the furthest to the left since the 75th Senate during the Great Depression and the New Deal. Further, the mean winning coordinate in the 113th House was the most conservative since the late nineteenth and early twentieth centuries, following the realigning election of 1896, after which Republicans controlled the House for 32 of the next 38 years. Conversely, the mean winning coordinate in the 113th Senate was the most liberal since the late nineteenth century.

Click images to enlarge



An Update on the Presidential Square Wave

Below we plot the first dimension DW-NOMINATE Common Space scores of the presidents in the post-war period, which we refer to as the “presidential square wave” due to its shape. An ideological score is estimated for each president throughout the entirety of their tenure in office by scaling their “votes” on a subset of roll call on which they announce a position (measured using CQ Presidential Support Votes). Negative DW-NOMINATE scores indicate greater liberalism and positive scores indicate greater conservatism. The presidential scores are directly comparable across time and with members of Congress.

These presidential DW-NOMINATE scores are estimated using all available CQ presidential support roll calls through 2013. CQ does not issue all of its presidential support roll calls until the print version of its congressional roll call guide comes out, and so only a fraction of the 2014 votes are available.

Very little has changed from the last presidential square wave. President Obama fits the spatial model estimated by DW-NOMINATE extremely well, with over 95% of his “votes” correctly classified. Obama has moved very slightly leftward (-0.367) and is now just to the left of LBJ (-0.346) and right of Truman (-0.368), though this trio is virtually ideologically indistinguishable. President Eisenhower is the most moderate president (0.293) of the post-war era.

Among members of the 113th Congress, President Obama is very ideologically close to Representatives Stephen Lynch (D-MA) [-0.364] and Loretta Sanchez (D-CA) [-0.369] in the House, and Senators Patty Murray (D-WA) [-0.367] and Mark Udall (D-CO) [-0.359] in the Senate.

Click image to enlarge


House and Senate Polarization 1879 – 2014

Below we show the polarization of the Political Parties for the 1879 through 2014 period (46th to 113th Congresses). Polarization is measured by the distance between the means of the Democrat and Republican Parties on the first (Liberal vs. Conservative) DW-NOMINATE dimension. Polarization is now at a Post-Reconstruction high in both the House and Senate.

Click image to enlarge


Below are the Party means for the House on the first DW-NOMINATE dimension. For the Democrats we show the means for the Northern and Southern wings of the Party (We use the CQ definition of South; the 11 States of the Confederacy plus Kentucky and Oklahoma). In the past three Congresses the difference between the Northern and Southern Democrats has disappeared.

Click image to enlarge


Below are the corresponding Party means for the Senate on the first DW-NOMINATE dimension. The pattern is essentially the same as the House. However, the Southern Democratic Senators as a group tend to be more moderate than their Northern counterparts.

Click image to enlarge


Below are the Party mean graphs for the House and second for the second DW-NOMINATE dimension. This dimension usually picks up regional differences between the two major parties. Before the Civil War the second dimension picked up the North vs. South division on Slavery. In the Post Reconstruction period the second dimension picked up regional differences on soft vs. hard Money (bimetalism, gold and silver) and beginning in the late 1930s Civil Rights. In the past 20 years the second dimension has faded in importance and issues that used to divide the parties internally — e.g., gun control, abortion — now load almost entirely on the first dimension. This is explored in detail in Keith T. Poole and Howard Rosenthal (2007) Ideology and Congress. Note that in the figures below the parties have almost converged on the second DW-NOMINATE dimension. Voting in Congress is almost entirely one-dimensional. The first dimension now accounts for over 93-94 percent of the roll call votes.

Click images to enlarge



House: Vote on $1.1 Trillion Spending Bill (the “Cromnibus”)

Below we use Optimal Classification (OC) in R to plot the House’s 219-206 vote on a $1.1 trillion spending package (the so-called Cromnibus). The vote split both party caucuses, with House Republicans voting 162-67 in favor of the bill and House Democrats splitting 57-139 against it.

The bill angered ideologues on both sides, and the result was a “two-ends-against-the-middle” vote in which a coalition of more moderate Republican and Democratic legislators united to pass the measure. The mean first dimension (representing liberal-conservative position) score of Democrats who voted Yea is -0.41, compared to a more liberal mean score of -0.49 for Nay Democrats (p < 0.01). The mean first dimension score of Yea Republicans is 0.38, compared to a more conservative mean score of 0.44 for Nay Republicans (p < 0.01).

Click image to enlarge


Note: The plot show only 217 Yea and 203 Nay votes because several newly-elected members have not yet cast enough votes to be included in the scaling.

The House and Senate Votes on the Keystone XL Pipeline

Below we use Optimal Classification (OC) in R to plot the House’s 252-161 vote and the Senate’s 59-41 vote on approving the Keystone XL pipeline. The Senate fell one vote short of approving the measure, which was designed in part to help Senator Mary Landrieu’s (D-LA) prospects in her upcoming runoff against Rep. Bill Cassidy (R-LA).

31 House Democrats and 14 Senate Democrats joined with all voting Republicans in both chambers in support of the measure. The first dimension (representing legislators’ liberal-conservative positions) does a good job at capturing this internal divide among House and (especially) Senate Democrats. Three of the Democratic Senators who voted Yea were defeated in their re-election races earlier this month — Senators Mark Begich (D-AK), Kay Hagan (D-NC), and Mark Pryor (D-AR) — and these are among the more moderate members of the Senate Democratic Caucus. This helps to illustrate why polarization is likely to increase in the next (114th) Senate.

Click images to enlarge